langextract-anthropic 0.2.0

LangExtract provider plugin for Anthropic Claude

LangExtract Anthropic Provider

A provider plugin for LangExtract that integrates Anthropic's Claude API for robust, structured information extraction.

Features

• Native Anthropic API: Uses the official anthropic Python SDK for Claude models.
• Safe parameter handling: Whitelist filtering; unsupported params raise clear errors.
• Concurrent batching: Parallel inference for multi-prompt workloads.
• Schema-aware: Optional structured output mode (JSON) from LangExtract examples.
• Modern packaging: pyproject.toml with Hatch; works well with uv.

Installation

Using UV (Recommended)

# Install UV if you haven't already
curl -LsSf https://astral.sh/uv/install.sh | sh

# Install the package
uv add langextract-anthropic

Using pip

pip install langextract-anthropic

From Source

git clone <repository-url>
cd langextract-anthropic
uv sync

Quick Start

1. Set up Anthropic API credentials

export ANTHROPIC_API_KEY="your-api-key"

2. Use with LangExtract

import langextract as lx

# Define extraction examples
examples = [
    lx.ExampleData(
        text="John Smith works at Microsoft in Seattle.",
        extractions=[
            lx.ExtractionData(
                extraction_class="Person",
                attributes={"name": "John Smith"}
            ),
            lx.ExtractionData(
                extraction_class="Organization", 
                attributes={"name": "Microsoft"}
            ),
            lx.ExtractionData(
                extraction_class="Location",
                attributes={"name": "Seattle"}
            ),
        ],
    ),
]

# Extract information using Anthropic Claude
result = lx.extract(
    text_or_documents="Sarah Johnson is a data scientist at Google in Mountain View.",
    prompt_description="Extract people, organizations, and locations.",
    examples=examples,
    model_id="anthropic-claude-3-5-sonnet-latest",
    temperature=0.1,
    max_tokens=512,
)

print(result.extractions)

Supported Models

This provider supports all Anthropic Claude models:

• claude-3-5-sonnet-latest (recommended)
• claude-3-5-sonnet-20241022
• claude-3-5-haiku-latest
• claude-3-opus-latest
• claude-3-sonnet-20240229
• claude-3-haiku-20240307

Model ID Format

Use the anthropic- prefix or specify the model name directly:

• anthropic-claude-3-5-sonnet-latest → Uses model: claude-3-5-sonnet-latest
• anthropic-claude-3-opus-latest → Uses model: claude-3-opus-latest
• claude-3-5-sonnet-latest → Uses model directly

Configuration Parameters

Core Parameters

Parameter	Type	Description	Default
model_id	str	Model identifier	"claude-3-5-sonnet-latest"
api_key	str	Anthropic API key	ANTHROPIC_API_KEY env var
temperature	float	Controls randomness (0-1)	None
max_workers	int	Parallel request workers	10

Anthropic API Parameters

Parameter	Type	Description	Range
max_tokens	int	Maximum tokens to generate	1-8192
temperature	float	Sampling temperature	0.0-1.0
top_p	float	Nucleus sampling	0.0-1.0
top_k	int	Top-k sampling	0-200
stop_sequences	list[str]	Stop sequences	Max 4 items
metadata	dict	Request tracking metadata	-

Usage Examples

# Basic extraction
result = lx.extract(
    text_or_documents=text,
    prompt_description=prompt,
    examples=examples,
    model_id="anthropic-claude-3-5-sonnet-latest",
)

# With custom parameters
result = lx.extract(
    text_or_documents=text,
    prompt_description=prompt,
    examples=examples,
    model_id="anthropic-claude-3-5-sonnet-latest",
    temperature=0.3,
    max_tokens=1000,
    top_p=0.9,
    stop_sequences=["END", "STOP"],
    metadata={"user_id": "user123"},
)

Environment Variables

Variable	Description	Required
ANTHROPIC_API_KEY	Anthropic API key	Yes

Development

Setup Development Environment

# Clone the repository
git clone <repository-url>
cd langextract-anthropic

# Install UV
curl -LsSf https://astral.sh/uv/install.sh | sh

# Install dependencies
uv sync --dev

Running Tests

# Run unit tests (no API calls)
uv run pytest tests/ -m "unit"

# Run integration tests (requires ANTHROPIC_API_KEY)
uv run pytest tests/ -m "integration" 

# Run all tests with coverage
uv run pytest tests/ --cov=langextract_anthropic --cov-report=html

Development Commands

# Format code
uv run black langextract_anthropic tests
uv run isort langextract_anthropic tests

# Lint code
uv run ruff check langextract_anthropic tests
uv run mypy langextract_anthropic

# Build package
uv build

# Bump version
python scripts/bump_version.py patch  # 0.1.0 -> 0.1.1
python scripts/bump_version.py minor  # 0.1.0 -> 0.2.0
python scripts/bump_version.py major  # 0.1.0 -> 1.0.0

Testing

This provider includes comprehensive testing:

• Unit tests: Mock-based testing of provider logic
• Parameter tests: Validation of API parameter filtering
• Integration tests: Real API testing (requires credentials)

# Set up test environment
export ANTHROPIC_API_KEY="your-api-key"

# Run specific test categories
uv run pytest tests/test_provider_unit.py -v
uv run pytest tests/test_parameter_filtering.py -v
uv run pytest tests/test_anthropic_integration.py -v  # requires API key

Error Handling

The provider provides clear error messages for common issues:

try:
    result = lx.extract(...)
except lx.exceptions.InferenceConfigError as e:
    # Configuration errors (missing API key, invalid params)
    print(f"Configuration error: {e}")
except lx.exceptions.InferenceRuntimeError as e:
    # Runtime errors (API failures, network issues)
    print(f"Runtime error: {e}")
    print(f"Original error: {e.original}")

Troubleshooting

Common Issues

1. Missing API Key

   InferenceConfigError: Anthropic API key not provided
   

   Solution: Set ANTHROPIC_API_KEY environment variable or pass api_key parameter.

2. Invalid Model Name

   AnthropicAPIError: model not found
   

   Solution: Use a valid Claude model name (see supported models above).

3. Rate Limiting

   AnthropicAPIError: 429 Too Many Requests
   

   Solution: Reduce max_workers or add retry logic in your application.

4. Token Limit Exceeded

   AnthropicAPIError: maximum context length exceeded
   

   Solution: Reduce input text length or increase max_tokens parameter.

License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request. For major changes, please open an issue first to discuss what you would like to change.

Changelog

See CHANGELOG.md for a list of changes and version history.